This study aimed to analyze the effect of exercise-induced hyperpnea on gross efficiency (GE) and anaerobic capacity estimates during a self-paced 3-min supramaximal cycle time trial (TT). Fourteen highly trained male cyclists performed 7 × 4-min submaximal stages, a 6-min passive rest, a 3-min TT, a 5-min passive rest, and a 6-min submaximal stage. Three models were based on the 7 × 4-min linear regression extrapolation method, using 1) the conventional model (7-YLIN); 2) the same 7-YLIN model but correcting for the additional ventilatory cost (i.e., hyperpnea) (7-YLIN-V-cor); and 3) accounting for linearly declining GE during the TT (7-YLIN-D). The other three models were based on GE from the last submaximal stage, using the conventional model (GELAST) and the same modifications as described for 7+YLIN, i.e., 1) GELAST, 2) GELAST-V-cor, and 3) GELAST-D. The GELAST model generated 18% higher values of anaerobic capacity than the 7-YLIN model (P < 0.05). During the TT, the hyperpnea-corrected model (i.e., 7-YLIN-V-cor or GELAST-V-cor) generated, compared with the respective conventional model (i.e., 7-YLIN or GELAST), ∼0.7 percentage points lower GE and ∼11% higher anaerobic capacity (all, P < 0.05). The post-TT GE was 1.9 percentage points lower (P < 0.001) and the 7-YLIN-D or GELAST-D model generated, compared with the respective conventional model, a lower GE (∼1.0 percentage points) and ∼17% higher anaerobic capacity during the TT (all, P < 0.05). In conclusion, the correction for a declining GE due to hyperpnea during a supramaximal TT resulted in an increased required total metabolic rate and anaerobic energy expenditure compared with the conventional models.NEW & NOTEWORTHY This study demonstrates that GE declines during a 3-min supramaximal cycle TT, which is possibly related to the hyperpneic response during supramaximal exercise. The finding from this study also provides novel insight into how the increased ventilatory energy cost from exercise-induced hyperpnea contributes to decreased GE, increased required total metabolic rate, and increased anaerobic energy expenditure during supramaximal exercise. Therefore, conventional linear models for estimating anaerobic capacity are likely to generate underestimated values.